Seeinstructions): Dynamic contrast-enhanced (DCE) magnetic resonance imaging (MRI) of breast cancer patients has shown considerable promise in aiding diagnoses of breast lesions and characterizing treatment response. The challenge in DCE breast imaging is the need for both good temporal resolution to capture tracer kinetic properties and good spatial resolution for visualizing morphology. Traditional dynamic methods in MRI acquire incomplete k-space data at each time point, and use k-space temporal interpolation (or data sharing) to form "complete" k-space datasets prior to Fourier reconstruction. We propose to investigate model-based image reconstruction methods that avoid k-space interpolation by estimating the object model parameters that best fit the available k-space data. These reconstruction methods will incorporate parallel imaging techniques. They will also be extended to account for nonrigid deformations due to patient motion during the scan using novel methods for joint estimation of motion parameters and image intensity parameters. The methods will be evaluated using computer simulations, phantom studies, and human DCE-MRI scan data. The human data will be collected as part of Project 1 and will include DCE-MRI scans of breast cancer patients undergoing neoadjuvant chemotherapy, where early prediction of tumor response is of clinical importance. The proposed methods have the potential to improve image quality both in breast DCE-MRI as well as other dynamic MR applications. RELEVANCE (See instructions): The relevance of this research to public health is that improving the quality of MR images through more sophisticated data processing may lead to more accurate diagnosis and treatment of patients with breast cancer and other diseases.